# E129 Forensic Baseline Audit & E128 Reconciliation

**Entity:** FIRME AI CORPORATION · **System:** CAIN-42 · **Evolution:** E129
**Generated:** 2026-11-21T18:47:01Z · **HEAD:** `cc9c9702d`
**AGI Status:** NOT ESTABLISHED · **RSI Status:** EXPERIMENTAL · **Baseline Check:** PASS
**Signed Manifest SHA-256:** `0bc9376bf9fcbf23eb5699985256809c97c37b11b3636162235b62013f8a3209`

## 1. Historical Baselines & Dual-Laboratory Verification

| Suite / Laboratory | Result | Summary |
|---|---|---|
| `e128` Offline Verifier | PASS | 61/61 checks passed; 0 failed |
| `e127` Offline Verifier | PASS | 89/89 checks passed; 0 failed |
| `e126b` Offline Verifier | PASS | 57/57 checks passed; 0 failed |
| `e126a` Offline Verifier | PASS | 140/140 checks passed; 0 failed |
| `e126` Offline Verifier | PASS | 66/66 checks passed; 0 failed |
| `rsi_e129_prior` Offline Verifier | PASS | } |
| Pytest Regression (E126–E128 + RSI E129) | PASS | ============================= 54 passed in 20.09s ============================== |

## 2. E128 Reconciliation (G24, G25, G26 & Live Deployment)

- **E128 Merkle Root:** `7bb44c0c03c5948f45d7ab0902a04b630361508c05b30f7aa73a4e3650d9a328` (`2,000` valid trials, `64` local Ollama LLM trials, `$0.00` paid).
- **Working-tree state at E129 start:** E128 implementation existed verified on disk (61/61 offline checks, 39/39 pytest) and was uncommitted on branch `evolution-79` above `cbb7022cf`.
- **G24 (Independent Reproduction):** Remains `PARTIAL` (`7` internal Worker-8 audits, `0` external third-party submissions).
- **G25 (External Benchmarks):** Remains `BLOCKED` (official ARC-AGI-2/3, SWE-bench Verified official, METR/HCAST, AI4AI-Bench unavailable).
- **G26 (Live Deployment Reconciliation):** All three static homepages and E128 pages returned HTTP 200 live, **but** `cain-platform-gateway.service` (last started `17:18:04Z`, 1 second prior to the E128 `main.py` router mount) served E127 `/api/v1/research/status` (`200`) while returning `404` on `/api/v1/research/health`. Hugging Face Space (`integrations/huggingface-space/README.md`) contains E127/dual-lab updates (`mentions_e127=True`) and does **not** yet contain E128 (`mentions_e128=False`).

### Pre-E129 Live Endpoint Probes

| URL | HTTP Status |
|---|---:|
| `https://clawx.click/` | 200 |
| `https://clawx.click/e128-observatory.html` | 200 |
| `https://clawx.click/api/v1/research/status` | 200 |
| `https://clawx.click/api/v1/research/health` | 404 |
| `https://cainstudio.online/` | 200 |
| `https://cainstudio.online/e128-workspace.html` | 200 |
| `https://cainstudio.online/api/v1/research/health` | 404 |
| `https://mcpgate.online/` | 200 |
| `https://mcpgate.online/e128-governance.html` | 200 |
| `https://mcpgate.online/api/v1/research/health` | 404 |

## 3. Preserved Negative Results & Honest Limitations

- E128 Causal Reasoning Active-vs-Random FAIL: active EIG MAE 0.7546 vs random 0.7055 (wins 111/131, Wilcoxon p=0.09303028956).
- E128 Local 0.5B-1.5B Ollama LLM accuracy: 18/64 (28.1%) across all 8 workers (W1 0.125, W2 0.50, W3 0.00, W4 0.00, W5 0.25, W6 0.625, W7 0.00, W8 0.75).
- CPU-only host limitations (8 vCPUs, 30 GiB RAM, no GPU; small quantized 0.5B-1.5B local models only; $0.00 paid spend).
- Single-host SQLite WAL scheduler (durable locally; NOT distributed and NOT Byzantine fault tolerant).
- External benchmarks (ARC-AGI-2, ARC-AGI-3, SWE-bench Verified official, METR/HCAST, AI4AI-Bench) remain BLOCKED_UNAVAILABLE.
- Independent third-party verification / external researcher reproductions remain 0 (OPEN).
- E128 G24 PARTIAL (0 external reproductions), G25 BLOCKED (external benchmarks), G26 PARTIAL (live gateway running since 17:18:04Z loaded pre-E128 E127 router so /api/v1/research/health returned 404 prior to E129 reconciliation; HF Space reflects E127/dual-lab, not E128).
- Prior Lab B RSI limitations preserved: E129 RSI bounded to constructed integer-rule DSL; E130 negative/partial; E133 tier-8 MITM reversal diagnosis (0/10 query score, NOT cited as tier-8 capability).
